Agent

This track holds papers, experiments, and benchmarks about agent roles, delegation, critique, recovery, and collaboration. It is for research notes and evaluation assets, not runtime code.